Knowledge of the domain of applicability of a machine learning model is essential to ensuring accurate and reliable model predictions.In this work,we develop a new and general approach of assessing model domain and de...Knowledge of the domain of applicability of a machine learning model is essential to ensuring accurate and reliable model predictions.In this work,we develop a new and general approach of assessing model domain and demonstrate that our approach provides accurate and meaningful domain designation across multiple model types and material property data sets.Our approach assesses the distance between data in feature space using kernel density estimation,where this distance provides an effective tool for domain determination.We show that chemical groups considered unrelated based on chemical knowledge exhibit significant dissimilarities by our measure.We also show that high measures of dissimilarity are associated with poor model performance(i.e.,high residual magnitudes)and poor estimates of model uncertainty(i.e.,unreliable uncertainty estimation).Automated tools are provided to enable researchers to establish acceptable dissimilarity thresholds to identify whether new predictions of their own machine learning models are in-domain versus out-of-domain.展开更多
We have developed a machine learning model for critical cooling rates for metallic glasses based on computational properties,supporting in-silico screening for desired Rc values and significantly reducing reliance on ...We have developed a machine learning model for critical cooling rates for metallic glasses based on computational properties,supporting in-silico screening for desired Rc values and significantly reducing reliance on time-consuming laboratory work.We compare results for features derived from easy-tocompute functions of elemental properties to more complex physically motivated properties using ab initio,machine-learning potentials,and empirical potential molecular dynamics methods.The established approach enables property acquisition across a diverse range of alloys.Analysis of various features for 34 alloys from 20 chemical systems shows that the best model for critical cooling rates was learned from one elemental property-based feature and three simulated features.The elemental property based feature is an ideal entropy value based on alloy stoichiometry.The simulated features were acquired from estimates of energies above the convex hull,changes in heat capacity,and the fraction of icosahedra-like Voronoi polyhedra.Models were assessed through a demanding cross validation test based on repeatedly leaving out full chemical systems as test sets and had an R2 of 0.78 and a mean average error of 0.76 in units of lg(K/s).We demonstrate with Shapley additive explanation analysis that the most impactful features have physically reasonable influence on model predictions.The established methodology can be applied to other high-throughput studies of material properties of diverse compositions.展开更多
基金the Bridge to the Doctorate:Wisconsin Louis Stokes Alliance for Minority Participation National Science Foundation(NSF)award number HRD-1612530the University of Wisconsin-Madison Graduate Engineering Research Scholars(GERS)fellowship program,and the PPG Coating Innovation Center for financial support for the initial part of this work.The other authors gratefully acknowledge support from the NSF Collaborative Research:Framework:Machine Learning Materials Innovation Infrastructure award number 1931306+1 种基金Lane E.Schultz also acknowledges this award for support for the latter part of this work.Machine learning was performed with the computational resources provided by XSEDE 2.0:Integrating,Enabling and Enhancing National Cyberinfrastructure with Expanding Community Involvement Grant ACI-1548562We thank former and current members of the Informatics Skunkworks at the University of Wisconsin-Madison for their contributions to early aspects of this work:Angelo Cortez,Evelin Yin,Jodie Felice Ritchie,Stanley Tzeng,Avi Sharma,Linxiu Zeng,and Vidit Agrawal.
文摘Knowledge of the domain of applicability of a machine learning model is essential to ensuring accurate and reliable model predictions.In this work,we develop a new and general approach of assessing model domain and demonstrate that our approach provides accurate and meaningful domain designation across multiple model types and material property data sets.Our approach assesses the distance between data in feature space using kernel density estimation,where this distance provides an effective tool for domain determination.We show that chemical groups considered unrelated based on chemical knowledge exhibit significant dissimilarities by our measure.We also show that high measures of dissimilarity are associated with poor model performance(i.e.,high residual magnitudes)and poor estimates of model uncertainty(i.e.,unreliable uncertainty estimation).Automated tools are provided to enable researchers to establish acceptable dissimilarity thresholds to identify whether new predictions of their own machine learning models are in-domain versus out-of-domain.
基金Wisconsin Louis Stokes Alliance for Minority Participation National Science Foundation(NSF)award number HRD-1612530the University of Wisconsine Madison Graduate Engineering Research Scholars(GERS)fellowship program for the financial support for graduate student investigation,and the PPG Coating Innovation Center for financial support.Dr.Afflerbach gratefully acknowledges research support from the NSF through the University of Wisconsin Materials Research Science and Engineering Center(DMR-2309000)+1 种基金All authors gratefully acknowledge support from the NSF Collaborative Research:Framework:Machine Learning Materials Innovation Infrastructure award number 1931306Machine learning was performed with the computational resources provided by XSEDE 2.0:Integrating,Enabling and Enhancing National Cyberinfrastructure with Expanding Community Involvement Grant ACI-1548562.
文摘We have developed a machine learning model for critical cooling rates for metallic glasses based on computational properties,supporting in-silico screening for desired Rc values and significantly reducing reliance on time-consuming laboratory work.We compare results for features derived from easy-tocompute functions of elemental properties to more complex physically motivated properties using ab initio,machine-learning potentials,and empirical potential molecular dynamics methods.The established approach enables property acquisition across a diverse range of alloys.Analysis of various features for 34 alloys from 20 chemical systems shows that the best model for critical cooling rates was learned from one elemental property-based feature and three simulated features.The elemental property based feature is an ideal entropy value based on alloy stoichiometry.The simulated features were acquired from estimates of energies above the convex hull,changes in heat capacity,and the fraction of icosahedra-like Voronoi polyhedra.Models were assessed through a demanding cross validation test based on repeatedly leaving out full chemical systems as test sets and had an R2 of 0.78 and a mean average error of 0.76 in units of lg(K/s).We demonstrate with Shapley additive explanation analysis that the most impactful features have physically reasonable influence on model predictions.The established methodology can be applied to other high-throughput studies of material properties of diverse compositions.
